Kafka Event Streaming
Explore three partition logs, two independent consumer groups and three replicas. Change membership, commit offsets, restart consumers and stop brokers to understand how Kafka handles state and failure.
Inside an event log
orders · 3 partitions · replication factor 3
Same key → same partition.
Committed lag 6 · Redeliveries 0
Committed lag 6 · Redeliveries 0
Six events are stored. Poll billing to see each partition advance independently.
Try an experiment · model boundaries
- Poll billing, crash before committing, then poll again. Watch redeliveries increase.
- Commit, restart and poll. Previously checkpointed records are skipped.
- Switch to analytics: it has its own offsets over the same retained log.
- Use 4 consumers: one is idle. Stop 2 brokers: publishing is rejected.
This is a teaching simulation, not a Kafka connection. It uses a simple key hash, round-robin ownership, immediate replication/elections and offset-based progress. Real partitioners, rebalances, ISR membership, transactions and failure timing are more complex. No data is saved.
Read the complete Kafka lessonRuns entirely in your browser — your data never leaves this page.